Haris Themistocleous

Journal Articles

1. Themistocleous, C., & Stark B. (2026). Language biomarker screening using AI: a transdiagnostic approach to the brain. Scientific Reports, 16, 4207. https://doi.org/10.1038/s41598-025-34257-z
2. Andreou, M., Themistocleous, C., & Peristeri, E. (2026). Integrating artificial intelligence and natural language processing to investigate lexical errors in autistic preschoolers. Research in Autism, 130, 202771. https://doi.org/https://doi.org/10.1016/j.reia.2025.202771
3. Tippett, D. C., Surrao, K., Neophytou, K., Kim, H., Gallegos, J., Themistocleous, C., Rapp, B., Hillis, A. E., & Tsapkini, K. (2025). Written picture descriptions distinguish variants of primary progressive aphasia. Journal of Alzheimer’s Disease, 13872877251376381. https://doi.org/10.1177/13872877251376381
4. Themistocleous Charalambos (2025). Linguistic and Emotional Prosody: A Systematic Review and ALE Meta-Analysis. Neuroscience & Biobehavioral Reviews, 106210. https://doi.org/https://doi.org/10.1016/j.neubiorev.2025.106210.
5. Themistocleous Charalambos (2024). Open Brain AI and language assessment. Frontiers in Human Neuroscience, 18. https://doi.org/10.3389/fnhum.2024.1421435
6. Kim, Hana., Hillis, Argye E., & Themistocleous Charalambos (2024). Machine Learning Classification of Patients with Amnestic Mild Cognitive Impairment and Non-Amnestic Mild Cognitive Impairment from Written Picture Description Tasks. Brain Sciences, 14(7).https://mdpi-res.com/d_attachment/brainsci/brainsci-14-00652/article_deploy/brainsci-14-00652.pdf?version=1719476290
7. Themistocleous Charalambos, Andreou Maria, Peristeri Eleni (2024). Autism detection in children: Integrating Machine Learning and Natural Language Processing in Narrative Analysis. Behavioral Sciences, 14(6), 459-475; https://doi.org/10.3390/bs14060459
8. Stark, Brielle C., Bryant, Lucy, Themistocleous Charalambos, den Ouden, Dirk - Bart, & Roberts, Angela C. (2023). Best practice guidelines for reporting spoken discourse in aphasia and neurogenic communication disorders. Aphasiology, 1 - 24. https://doi.org/10.1080/02687038.2022.2039372
9. Themistocleous, Charalambos, Fyndanis, Valantis, & Tsapkini, Kyrana. (2022). Sonorant spectra and coarticulation distinguish speakers with different dialects. Speech Communication.https://doi.org/10.1016/j.specom.2022.06.002.
10. Afthinos Afthinos, Themistocleous Charalambos,Herrmann Olivia, Fan Hongli, Lu Hanzhang, Tsapkini, Kyrana (2022). The Contribution of Working Memory Areas to Verbal Learning and Recall in Primary Progressive Aphasia. Frontiers in Neurology, 13. https://doi.org/10.3389/fneur.2022.698200.
11. Stark Brielle C., Bryant Lucy, Themistocleous Charalambos, den Ouden Dirk-Bart, Roberts Angela C. (2022). Best practice guidelines for reporting spoken discourse in aphasia and neurogenic communication disorders. Aphasiology, 1-24. https://doi.org/10.1080/02687038.2022.2039372.
12. Themistocleous Charalambos, Webster Kimberly, Tsapkini, Kyrana (2021). Effects of tDCS on Sound Duration in Patients with Apraxia of Speech in Primary Progressive Aphasia. Themistocleous Charalambos, Webster Kimberly, Tsapkini, Kyrana (2021). Effects of tDCS on Sound Duration in Patients with Apraxia of Speech in Primary Progressive Aphasia. Brain Sciences, 11(3). https://doi.org/10.3390/brainsci11030335.
13. Themistocleous Charalambos, Bronte Ficek, Kimberly Webster, Dirk-Bart den Ouden, Argye E. Hillis, Kyrana Tsapkini (2021). Automatic subtyping of individuals with Primary Progressive Aphasia. Journal of Alzheimer’s Disease. https://doi.org/10.3233/JAD-201101.
14. Themistocleous Charalambos, Kyriaci Neophytou, Brenda Rapp, Tsapkini Kyrana (2020). Au- tomatic metrics of spelling errors on words and non-words. Journal of Speech, Language, and Hearing Research. https://doi.org/10.1044/2020_JSLHR-20-00177.
15. Themistocleous Charalambos, Webster Kimberly, Alexandros Afthinos, and Tsapkini Kyrana (2020). Morphological markers in speech production distinguish individuals with different primary progressive aphasia variants: A natural language processing and machine learning approach. American Journal of Speech-Language Pathology. https://pubs.asha.org/doi/10.1044/2020_AJSLP-19-00114.
16. Themistocleous, Charalambos, Eckerström Marie, Kokkinakis Dimitrios (2020). Voice quality and speech fluency distinguish individuals with Mild Cognitive Impairment from Healthy Controls. PLOS ONE 5(7): e0236009. https://doi.org/10.1371/journal.pone.0236009.
17. Stark Brielle C., Dutta Manaswita, Murray Laura, Bryant Lucy, Fromm Davida; MacWhinney Brian, Ramage Amy, Simmons-Mackie Nina, Roberts Angela, Fergadiotis Gerasimos, den Ouden Dirk, Brock Kris, McKinney-Bock Katy, Jin Paek Eun, Harmon Tyson, Themistocleous Charalambos, Yoo Hyunsoo, On Yoon Si, Sharma Saryu, Aveni Katharine, Gutierrez Stephanie, Doppelbauer Lea, Dawn Bunker Lisa (2020). Standardizing assessment of spoken discourse in aphasia: A working group with deliverables. American Journal of Speech-Language Pathology. 1–12.
18. Georgiou George and Themistocleous Charalambos (2020). Vowel learning in diglossic settings: Evidence from Arabic-Greek learners. International Journal of Bilingualism. https://doi.org/10.1177/1367006920945396.
19. Themistocleous Charalambos (2019). Dialect classification from a single sonorant sound using Deep Neural Networks. Frontiers in Communication. https://doi.org/10.3389/fcomm.2019.00064.
20. Fyndanis Valantis and Themistocleous Charalambos (2019). Are there prototypical associations between time frames and aspectual values? Evidence from Greek aphasia and healthy ageing, Clinical Linguistics and Phonetics.
21. Themistocleous Charalambos, Eckerström Marie, and Dimitrios Kokkinakis (2018). Identification of Mild Cognitive Impairment from Speech in Swedish using Deep Sequential Neural Networks. Frontiers in Neurology 9, 975. URL: https://www.frontiersin.org/article/10.3389/fneur.2018.00975.
22. Themistocleous, Charalambos (2017). Effects of two linguistically proximal varieties on the spectral and coarticulatory properties of fricatives: Evidence from Athenian Greek and Cypriot Greek. Frontiers in Psychology.
23. Themistocleous, Charalambos (2017). Dialect classification using vowel acoustic parameters. Speech Communication 94, 13 – 22.
24. Grohmann Kleanthes, Papadopoulou Elena and Themistocleous Charalambos (2017). Acquiring Clitic Placement in Bilectal Settings: Interactions between Social Factors. Frontiers in Communication 2:5.
25. Themistocleous, Charalambos (2017). The Nature of Phonetic Gradience across a Dialect Continuum: Evidence from Modern Greek Vowels. Phonetica 74, 157–172.
26. Themistocleous, Charalambos (2016). Seeking an anchorage: Evidence from the tonal alignment of the Cypriot Greek prenuclear pitch accent. Language and Speech, 59(4):433–461.
27. Themistocleous, Charalambos (2016). The bursts of stops can convey dialectal information. Journal of the Acoustical Society of America EL 140(4), EL334–EL340.
28. Themistocleous, Charalambos (2014). Edge-Tone Effects and Prosodic Domain Effects on Final Lengthening. Linguistic Variation 14(1). 129–160. 

Books

1. Themistocleous Charalambos (2015). Experimental Design and Statistics in Linguistics. Using R (In Greek). Athens: Ion. ISBN:978-960-508-187-4.
2. Fragaki Georgia, Georgakopoulos Athanasios, Themistocleous Charalambos (Eds.) (2012). Current Trends in Greek Linguistics. Newcastle upon Tyne: Cambridge Scholars Publishing.

• Themistocleous, Charalambos & Stark, Brielle C. (2026). Language biomarker screening using AI: a transdiagnostic approach to the brain. Scientific Reports. 16. doi: 10.1038/s41598-025-34257-z. Full text in Research Archive Show summary

  Individuals with left-hemisphere damage (LHD), right-hemisphere damage (RHD), dementia, mild cognitive impairment (MCI), traumatic brain injury (TBI), and healthy controls are characterized by overlapping clinical profiles affecting communication and social interaction. Language provides a rich, non-invasive window into neurological health, yet objective and scalable methods to automatically differentiate between conditions with are lacking. This method aims to develop comprehensive neurolinguistic measures of these conditions, develop a machine learning multiclass screening and language assessment model (NeuroScreen) and offer a large comparative database of measures for other studies to build upon. We combined one of the largest databases, comprising 291 linguistic biomarkers calculated from speech samples produced by 1,394 participants: 536 individuals with aphasia secondary to LHD, 193 individuals with dementia, 107 individuals with MCI, 38 individuals with RHD, 58 individuals with TBI, and 498 Healthy Controls. Employing natural language processing (NLP) via the Open Brain AI platform (http://openbrainai.com), we extracted multiple linguistic features from the speech samples, including readability, lexical richness, phonology, morphology, syntax, and semantics. A Deep Neural Network architecture (DNN) classifies these conditions from linguistic features with high accuracy (up to 91%). A linear mixed-effects model approach was employed to determine the biomarkers of the neurological conditions, revealing distinct, quantitative neurolinguistic properties: LHD and TBI show widespread deficits in syntax and phonology; MCI is characterized by fine-grained simplification; patients with dementia present with specific lexico-semantic impairments; and RHD shows the most preserved profile. Ultimately, the outcomes provide an automatic detection and classification model of key neurological conditions affecting language, along with a novel set of validated neurological markers for facilitating differential diagnosis, remote monitoring, and personalized neurological care.

• Stark, Brielle C. & Themistocleous, Charalambos (2026). Normative clinical language data and task specific effects. Brain and Language. ISSN 0093-934X. 283. doi: 10.1016/j.bandl.2026.105850. Full text in Research Archive
• Sundström, Simon & Themistokleous, Charalampos (2026). A whole-word computational algorithm for phonological error scoring: Differentiating children with developmental language disorder, hearing impairment, and typical development. Clinical Linguistics & Phonetics. ISSN 0269-9206. p. 1–16. doi: 10.1080/02699206.2026.2704602. Full text in Research Archive
• Andreou, Maria; Themistokleous, Charalampos & Peristeri, Eleni (2025). Integrating artificial intelligence and natural language processing to investigate lexical errors in autistic preschoolers. Research in Autism Spectrum Disorders. ISSN 1750-9467. 130. doi: 10.1016/j.reia.2025.202771. Full text in Research Archive Show summary

  Expressive vocabulary has been characterized as a strength in the autistic population, with most studies focusing on the cues that autistic children use to learn new words. There is little knowledge about the organization of lexical networks in autism. This study aimed to characterize the lexical retrieval errors in a cohort of 60 Greek-speaking autistic preschoolers by combining manual coding with distributional semantics. Children completed a standardized picture-naming task yielding 745 errors classified into four error types: semantic, visual, phonological, and irrelevant. To quantify each error’s proximity to its intended target, we applied a pre-trained word embeddings model to derive continuous semantic proximity scores. More than half of the children exhibited expressive vocabulary abilities which were well below their chronological age level, while the opposite pattern was observed for one third of the children. Within this sample, semantic errors were predominant, and word embedding-derived distances corroborated the manual taxonomy: semantic substitutions displayed the highest proximity to target concepts, followed in descending order by phonological, visual, and irrelevant errors. Semantic errors showed no association with vocabulary delay, whereas irrelevant and visual errors increased with vocabulary delay, suggesting reduced semantic constraints in children whose vocabulary age equivalents were lower than their chronological ages. These findings underscore that naming errors in autistic children are not random but organized along graded semantic relationships, paving the way for computational modeling of lexical networks in this population. Integrating distributional-semantic modeling with traditional error analysis yields novel quantitative metrics for distinguishing between lexical error types, and paves the way for investigations into the development of lexical networks in autism.

• Themistokleous, Charalampos (2025). Linguistic and emotional prosody: A systematic review and ALE meta-analysis. Neuroscience and Biobehavioral Reviews. ISSN 0149-7634. 175. doi: 10.1016/j.neubiorev.2025.106210. Full text in Research Archive Show summary

  Prosody is a cover term referring to the melodic aspects of speech, with linguistic and affective (a.k.a. emotional) meanings. This review provides an overview of linguistic and affective prosody, evaluating two hypotheses on healthy individuals' linguistic and affective prosody. The first hypothesizes that the biological nature of affective prosody triggers activations unrelated to language (biological hypothesis), and the second that the aspects of affective prosody have been grammaticalized, i.e., incorporated into the language (linguistic hypothesis). We employed a systematic ALE metanalytic approach to identify neural correlates of prosody from the literature. Specifically, we assessed papers that report brain coordinates from healthy individuals selected using systematic research from academic databases, such as PubMed (NLM), Scopus, and Web of Science. We found that affective and linguistic prosody activate bilateral frontotemporal regions, like the Superior Temporal Gyrus (STG). A key difference is that affective prosody involves subcortical structures like the amygdala, and linguistic prosody activates linguistic areas and brain areas of social cognition and engagement. The shared activations, therefore, suggest that linguistic and affective meanings are combined, involving shared underlying brain connectivity mechanisms and acoustic manifestations. We suggest that the traditional distinction between linguistic and affective prosody may be overly rigid. Much like speech, lexicon, and grammar—domains that convey affective, social, and linguistic meanings without explicitly reflecting separate categories—prosody, too, functions as a system interfacing with affective, social, and linguistic domains. We conclude by proposing a novel blending hypothesis: prosody should be viewed as an integrated system that serves both affective and linguistic functions.

• Kim, Hana; Hillis, Argye E. & Themistokleous, Charalampos (2024). Machine Learning Classification of Patients with Amnestic Mild Cognitive Impairment and Non-Amnestic Mild Cognitive Impairment from Written Picture Description Tasks. Brain Sciences. 14(7). doi: 10.3390/brainsci14070652. Full text in Research Archive Show summary

  Individuals with Mild Cognitive Impairment (MCI), a transitional stage between cognitively healthy aging and dementia, are characterized by subtle neurocognitive changes. Clinically, they can be grouped into two main variants, namely patients with amnestic MCI (aMCI) and non-amnestic MCI (naMCI). The distinction of the two variants is known to be clinically significant as they exhibit different progression rates to dementia. However, it has been particularly challenging to classify the two variants robustly. Recent research indicates that linguistic changes may manifest as one of the early indicators of pathology. Therefore, we focused on MCI’s discourse-level writing samples in this study. We hypothesized that a written picture description task can provide information that can be used as an ecological, cost-effective classification system between the two variants. We included one hundred sixty-nine individuals diagnosed with either aMCI or naMCI who received neurophysiological evaluations in addition to a short, written picture description task. Natural Language Processing (NLP) and a BERT pre-trained language model were utilized to analyze the writing samples. We showed that the written picture description task provided 90% overall classification accuracy for the best classification models, which performed better than cognitive measures. Written discourses analyzed by AI models can automatically assess individuals with aMCI and naMCI and facilitate diagnosis, prognosis, therapy planning, and evaluation.

• Themistokleous, Charalampos (2024). Open Brain AI and language assessment. Frontiers in Human Neuroscience. 18. doi: 10.3389/fnhum.2024.1421435. Full text in Research Archive Show summary

  Neurolinguistic assessments play a vital role in neurological examinations, revealing a wide range of language and communication impairments associated with developmental disorders and acquired neurological conditions. Yet, a thorough neurolinguistic assessment is time-consuming and laborious and takes valuable resources from other tasks. To empower clinicians, healthcare providers, and researchers, we have developed Open Brain AI (OBAI). The aim of this computational platform is twofold. First, it aims to provide advanced AI tools to facilitate spoken and written language analysis, automate the analysis process, and reduce the workload associated with time-consuming tasks. The platform currently incorporates multilingual tools for English, Danish, Dutch, Finnish, French, German, Greek, Italian, Norwegian, Polish, Portuguese, Romanian, Russian, Spanish, and Swedish. The tools involve models for (i) audio transcription, (ii) automatic translation, (iii) grammar error correction, (iv) transcription to the International Phonetic Alphabet, (v) readability scoring, (vi) phonology, morphology, syntax, semantic measures (e.g., counts and proportions), and lexical measures. Second, it aims to support clinicians in conducting their research and automating everyday tasks with “OBAI Companion,” an AI language assistant that facilitates language processing, such as structuring, summarizing, and editing texts. OBAI also provides tools for automating spelling and phonology scoring. This paper reviews OBAI’s underlying architectures and applications and shows how OBAI can help professionals focus on higher-value activities, such as therapeutic interventions.

• Themistokleous, Charalampos; Andreou, Maria & Peristeri, Eleni (2024). Autism Detection in Children: Integrating Machine Learning and Natural Language Processing in Narrative Analysis. Behavioral Sciences. 14(6). doi: 10.3390/bs14060459. Full text in Research Archive Show summary

  Despite the consensus that early identification leads to better outcomes for individuals with autism spectrum disorder (ASD), recent research reveals that the average age of diagnosis in the Greek population is approximately six years. However, this age of diagnosis is delayed by an additional two years for families from lower-income or minority backgrounds. These disparities result in adverse impacts on intervention outcomes, which are further burdened by the often time-consuming and labor-intensive language assessments for children with ASD. There is a crucial need for tools that increase access to early assessment and diagnosis that will be rigorous and objective. The current study leverages the capabilities of artificial intelligence to develop a reliable and practical model for distinguishing children with ASD from typically-developing peers based on their narrative and vocabulary skills. We applied natural language processing-based extraction techniques to automatically acquire language features (narrative and vocabulary skills) from storytelling in 68 children with ASD and 52 typically-developing children, and then trained machine learning models on the children’s combined narrative and expressive vocabulary data to generate behavioral targets that effectively differentiate ASD from typically-developing children. According to the findings, the model could distinguish ASD from typically-developing children, achieving an accuracy of 96%. Specifically, out of the models used, hist gradient boosting and XGBoost showed slightly superior performance compared to the decision trees and gradient boosting models, particularly regarding accuracy and F1 score. These results bode well for the deployment of machine learning technology for children with ASD, especially those with limited access to early identification services.

View all works in NVA